AMDZone has a statement from an AMD PR person about 90nm processors.
Let’s look at what it says, carefully:
“Our transition to 90nm for AMD64 processors is on track.”
Well, there are old tracks and new tracks. For instance, a few years back, we were supposed to see a 90nm Hammer, uhhh, late last year. In November 2002, AMD was telling analysts the same thing and said Dresden would be 100% 90nm by now (see page 12).
So AMD certainly got derailed from that track. What AMD has been saying since late last year is a new track which says, “production starts by the end of Q2 2004, and revenues from it by the end of Q3 2004.”
So the AMD statement is nothing new; it’s exactly what they’ve been saying the last nine months.
I don’t doubt AMD will manage to make 90nm chips in that timeframe. But that’s not the question, at least not in my mind.
The question is not, “Can you make 90nm chips shortly?” The question is “Can you make fast 90nm chips shortly?”
And when you’re not putting out a 2.6GHz 90nm chip in the fall like all your roadmaps indicated you were going to, and apparently not putting out a fast 90nm chip until sometime next year, your actions seem to indicate the answer is “No.”
To be, or not to be, that is NOT the question. To be fast, or not to be fast, is the question. 🙂
This is what people are questioning, and this statement is no answer.
Remember what happened with the original Hammers. AMD had no problem making 130nm Hammers that would work; they had a hell of a time getting them to work at competitive speeds.
When you read the rest of the comments, keep this in mind, and see if the AMD person says anything to indicate they’re having no problems making fast ones.
“We are having strong yields, we are having excellent power consumption, heat dissipation is is excellent as well.”
Again, for 90nm chips. I don’t doubt a 1.8 or 2.0GHz 90nm SOI chip would be all of the abovementioned. IBM has been making 90nm 2GHz SOI chips with no problem for a while now.
But what’s happening at 2.6 or 2.8?
“We will initiate volume 90nm production in the 2nd quarter, and we expect to deliver products for revenue in the 3rd quarter of 04. So we are on track.”
Again, this is just about 90nm processors, not fast ones, and this has been the standard message the last six months.
“The current 130nm AMD Athlon 64 FX core is capable of scaling for performance needs for customers for that segment for the remainder of the year.”
Wait a minute. If you have this wonderfully working 90nm process cranking out chips three months from now, why wouldn’t you use your latest and greatest process for your best desktop chip? Why would you instead stretch 130nm technology to the max instead when you have a better technology just sitting around waiting to be used, one that lets you makes them smaller, cheaper and more profitable?
If 90nm process technology were in good or even normal shape at 2.6GHz, yields at 90nm would almost certainly be much better than they’ll be at 130nm. It’s not like AMD sells millions of FXs; they ought to be able to make a few tens of thousands of FXs quite quickly.
Unless, of course, they’re having the same kind of problems Intel and IBM have been having.
Didn’t Intel Do The Same Thing With Prescott?
Why, yes, Intel played exactly this kind of word game with Prescott. They said everything was fine, the yields were good, they were on track to ramp up, etc., etc., etc. . . .
And what happened? Well, we found out Intel had absolutely no problem making slow Prescotts, but anything above that ranged from dicey to non-existent.
In other words, they could make them, but they couldn’t make fast ones, so Intel ripped up the roadmaps and suddenly decided Northwoods were OK enough for 3.4GHz.
Doesn’t this sound like the same old tune?
Fool Me Once, Shame On You. Fool Me Twice, Shame On Me. Fool Me More?
First, we had AMD denying problems with 130nm Hammers by saying yields were good, etc. when it was obvious to everyone in the world that they couldn’t make fast ones. So they delayed and pushed back desktop Hammers over six months until they could.
Next, we had Intel denying problems with 90nm Prescotts by saying yields were good, etc., then it became obvious to everyone in the world that they couldn’t make fast ones. So they delayed Prescott for a year, and now are delaying and pushing back any faster Prescotts, substituting Northwoods where they can.
After that, we had IBM promising a 3GHz 90nm SOI chip to Apple, but found itself unable to get more than 2.5GHz out the door in time. IBM isn’t delaying and pushing back 3GHz chips; they just said don’t expect to see one anytime soon.
Now we have AMD denying problems with 90nm Hammers by saying yields are good, etc., but delaying and pushing back the introduction of fast ones and substituting 130nm ones where they can.
How many times does this song have to be played before people catch on that it’s the same old tune?
A Few Simple Questions
If this is untrue or inaccurate, AMD can certainly resolve that by answering a few key questions. Questions like:
Is AMD currently able to make a sizable percentage (say, 25% or more) of 90nm Hammers able to run under default conditions at 2.6GHz or more?
When does AMD plan on releasing 2.6 and 2.8GHz 90nm Hammers?
At what speeds will the initial 90nm Hammers be rated?
If the initial fastest 90nm Hammers will be released at speeds lower than 2.6GHz, why is that? Would it not be more profitable to also sell higher-speed chips at higher prices?
If anybody thinks AMD is being done a bad turn by all this, they ought to ask AMD these questions to give them the opportunity to clear things up. If they won’t, you have to wonder why. I’d bet dollars to donuts, though, they’d refuse to answer them. After all, if everything were going great, that PR statement would have included wording indicating that making fast ones was no problem (plus, they wouldn’t be trying to stretch 130nm to 2.6GHz).